A dynamic reference index calculation method and system for grassland ecological quality evaluation

Abstract

The present application relates to the technical field of grassland ecological evaluation, and particularly relates to a dynamic reference index calculation method and system for grassland ecological quality evaluation. The method comprises the following steps: obtaining historical baseline data of a target region covering different degrees of disturbance and undegraded sample plots, and constructing a reference system time series prediction model based on the historical baseline data; obtaining a dynamic reference index potential value by using the reference system time series prediction model based on historical monitoring data of a grassland to be evaluated; extracting a disturbance factor from the historical baseline data, and setting a disturbance weight for the disturbance factor; and correcting the dynamic reference index potential value by using the disturbance weight based on the historical monitoring data and the historical baseline data to obtain a dynamic reference index. The present application solves the problem that the fixed reference index in the prior art cannot dynamically respond to changes in the environment, thereby resulting in inaccurate evaluation of the ecological condition of the grassland. The present application greatly improves the scientificity of the calculation of the reference index.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of grassland ecological evaluation, and particularly relates to a dynamic reference index calculation method and system for grassland ecological quality evaluation. BACKGROUND

[0002] As a core component of terrestrial ecosystems, the quality evaluation of grassland ecosystems is an important basis for ecological protection decision-making and sustainable management of grassland resources. Currently, the static reference index system widely used in grassland ecological quality evaluation usually takes the undegraded sample data of a certain specific historical period as a fixed benchmark to measure the ecological conditions of the target area.

[0003] However, this static reference mode has significant limitations: on the one hand, grassland ecosystems have dynamic evolution characteristics, and their structure and function are significantly affected by annual environmental conditions (such as precipitation distribution and temperature fluctuations), and the fixed benchmark is difficult to reflect the natural fluctuation rules of the ecological system in different years; on the other hand, the annual differences in disturbance factors such as grazing intensity and drought frequency have a significant impact on the quality of grassland ecosystems, and the static reference does not include quantitative analysis of such dynamic disturbances, resulting in a deviation between the evaluation results and the actual ecological conditions, making it difficult to accurately support the development of grassland protection and restoration measures. Therefore, it is urgent to develop a reference index calculation method that can dynamically respond to environmental changes and disturbance intensity to improve the scientificity and timeliness of grassland ecological quality evaluation. SUMMARY

[0004] In view of the defects in the prior art, the present application provides a dynamic reference index calculation method and system for grassland ecological quality evaluation, which solves the problem that the fixed reference index in the prior art cannot accurately evaluate the ecological conditions of the grassland.

[0005] In order to achieve the above-mentioned purpose, one aspect of the present application provides a dynamic reference index calculation method for grassland ecological quality evaluation, which comprises: obtaining historical baseline data of undegraded sample plots in a target area covering different disturbance levels, and constructing a reference system time series prediction model based on the historical baseline data; obtaining a dynamic reference index potential value by using the reference system time series prediction model based on the historical monitoring data of the grassland to be evaluated; extracting disturbance factors from the historical baseline data and setting disturbance weights for the disturbance factors; correcting the dynamic reference index potential value by using the disturbance weights based on the historical monitoring data and the historical baseline data to obtain a dynamic reference index.

[0006] This invention constructs a predictive model by combining historical data from non-degraded sample plots with different levels of disturbance, taking into account the temporal evolution of the ecosystem, introducing a disturbance factor weight correction mechanism, accurately quantifying the impact of disturbance, enabling the reference index to dynamically respond to environmental and disturbance changes, breaking through the limitations of traditional static benchmarks, and improving the scientificity and practicality of dynamic reference index calculation.

[0007] Optionally, obtaining historical baseline data of the target area covering non-degraded sample plots with different levels of interference includes: obtaining original historical data of the target area covering non-degraded sample plots with different levels of interference; imputing missing values ​​and removing outliers from the original historical data to obtain optimized historical data; and normalizing the optimized historical data to obtain historical baseline data.

[0008] This invention optimizes data quality by imputing missing values ​​and removing outliers, and then eliminates dimensional differences through normalization. This preserves the characteristics of the sample plot interference gradient and the integrity of ecological data, while ensuring data comparability. It provides a reliable foundation for the subsequent construction of a reference system time series prediction model and improves the integrity and accuracy of historical benchmark data.

[0009] Optionally, the historical baseline data includes grassland subtypes, historical biological indicators, historical environmental data, and historical disturbance data. The historical biological indicators include native functional group species cover, native functional group species aboveground biomass, degraded indicator species cover, and degraded indicator species aboveground biomass. The step of constructing a reference system time-series prediction model based on the historical baseline data includes: calculating the biomass proportion of degraded indicator species in the sample plot using the aboveground biomass of degraded indicator species; calculating the coverage proportion of degraded indicator species in the sample plot using the coverage of degraded indicator species; calculating the importance value of degraded indicator species in the sample plot based on the biomass proportion and the coverage proportion; constructing training samples based on the grassland subtype, native functional group species cover, native functional group species aboveground biomass, degraded indicator species importance value, historical environmental data, and historical disturbance data; constructing a reference system time-series prediction model; and training and evaluating the reference system time-series prediction model using the training samples.

[0010] This invention comprehensively reflects the ecological status of degraded indicator species in sample plots by calculating their biomass proportion, cover proportion, and importance value. It then combines grassland subtypes, native functional group indicators, and environmental and disturbance data to construct training samples, ensuring that the model input covers key ecosystem elements. The training samples take into account different grassland types and the synergistic relationships of biological indicators. The trained reference system time-series prediction model can accurately capture the temporal evolution patterns of ecological indicators, improving the ecological rationality and accuracy of dynamic reference indicator potential value prediction.

[0011] Optionally, the constructing the reference system time series prediction model comprises: performing statistical analysis on the historical benchmark data to obtain variation laws of native functional group coverage and native functional group species aboveground biomass; setting a synergistic constraint term based on the variation laws; adding the synergistic constraint term to a loss function constructed with mean square error to obtain an optimized loss function; and constructing the reference system time series prediction model according to the optimized loss function.

[0012] The application obtains variation laws of native functional group coverage and biomass by analyzing historical data, sets a synergistic constraint term based on the variation laws, and integrates the synergistic constraint term into a loss function to form an optimized loss function, so that the model can guarantee prediction accuracy and follow the ecological index synergistic evolution logic during training, avoid unreasonable results of contradictory coverage and biomass change trends, and improve the performance and practical application ability of the reference system time series prediction model.

[0013] Optionally, the extracting interference factors from the historical benchmark data and setting interference weights for the interference factors comprises: extracting historical benchmark interference data from the historical benchmark data, and determining interference factors according to the historical benchmark interference data; calculating a native functional group species annual comprehensive change rate by using the historical benchmark data; training a random forest change rate prediction model by using the historical benchmark interference data and the native functional group species annual comprehensive change rate; and calculating interference weights of the interference factors based on the random forest change rate prediction model.

[0014] The application determines interference factors by extracting interference data from historical benchmark data, trains a random forest model in combination with a native functional group species annual comprehensive change rate, accurately quantifies the weights of each interference factor, comprehensively covers natural and artificial interference types, integrates coverage and biomass dynamics through the comprehensive change rate, makes the interference weight calculation conform to the ecological response law, the random forest model can effectively capture the complex correlation between interference and functional group change, objectively gives each factor weight, avoids subjective weighting bias, and improves the scientificity and accuracy of weight calculation.

[0015] Optionally, the step of calculating the annual comprehensive change rate of native functional group species using the historical benchmark data includes: extracting the annual change rate set of native functional group species cover and the annual change rate set of native functional group species aboveground biomass from the historical benchmark interference data; calculating the mean, standard deviation, mean, and standard deviation of the annual change rate of cover, respectively, based on the annual change rate set of native functional group species cover and the annual change rate set of native functional group species aboveground biomass; calculating cover stability using the mean and standard deviation of the annual change rate of cover, and calculating biomass stability using the mean and standard deviation of the annual change rate of biomass; calculating the native functional group species cover weight and native functional group species aboveground biomass weight using the cover stability and biomass stability, respectively; and calculating the annual comprehensive change rate of native functional group species using the historical benchmark data based on the native functional group species cover weight and native functional group species aboveground biomass weight.

[0016] This invention extracts the annual change rate sets of cover and biomass step by step, calculates the mean and standard deviation to determine their stability, and then uses the stability to deduce the weights, finally integrating them to obtain the annual comprehensive change rate. It captures the interannual dynamic characteristics of the indicators through the mean and standard deviation, and quantifies the sensitivity of the indicators to disturbances through stability, making the weight allocation more in line with the ecological response law. The comprehensive change rate integrates the synergistic dynamics of cover and biomass, avoids the one-sidedness of a single indicator, accurately reflects the overall change trend of the original functional groups, and improves the accuracy of the calculation of the annual comprehensive change rate of the original functional groups.

[0017] Optionally, calculating the interference weight of the interference factor based on the random forest rate of change prediction model includes: extracting node information of the interference factor participating in node splitting in the random forest rate of change prediction model; calculating the mean squared error decrease value of the splitting node dominated by the interference factor based on the node information; calculating the total importance of the interference factor in the random forest model using the mean squared error decrease value; and calculating the interference weight of the interference factor based on the total importance.

[0018] This invention extracts information about interference factors participating in node splitting in a random forest model, calculates the mean squared error decrease value to quantify their ability to distinguish changes in functional groups, and then accumulates them to obtain the total importance and normalizes them to determine the weights. This approach not only objectively mines the actual impact of interference factors from the model's decision-making logic, but also accurately reflects their contribution to prediction accuracy through the mean squared error decrease value, avoiding subjective weighting bias and further improving the accuracy and scientific nature of interference weight calculation.

[0019] Optionally, the correcting the dynamic reference index potential value to obtain the dynamic reference index based on the historical monitoring data and the historical reference data using the interference weight comprises: calculating a current standardized intensity of the interference factor using the historical monitoring data and the historical reference data; calculating a current total influence amplitude of the interference using the current standardized intensity and the interference weight; and correcting the dynamic reference index potential value based on the current total influence amplitude of the interference to obtain the dynamic reference index.

[0020] The application eliminates the dimensional differences of different interference factors by normalization, and reflects the ecological influence of each factor by weight, so that the total influence amplitude can objectively reflect the comprehensive effect of the current interference. The corrected dynamic reference index integrates real-time interference information, breaks through the limitation of static reference, is more consistent with the actual state of the grassland ecosystem, and improves the timeliness and accuracy of the evaluation.

[0021] Optionally, the calculating the current standardized intensity of the interference factor using the historical monitoring data and the historical reference data comprises: extracting monitoring interference data of the last year in the historical monitoring data; calculating a historical interference mean and a historical interference standard deviation according to the historical reference data; and calculating the current standardized intensity of the interference factor based on the historical interference mean, the historical interference standard deviation and the monitoring interference data.

[0022] The application extracts the interference data of the last year in the historical monitoring data, combines the interference mean and the standard deviation of the historical reference data to obtain the current standardized intensity of the interference factor, focuses on the latest interference state, quantifies the degree of deviation of the current interference from the normal state with reference to the historical data, eliminates the dimensional differences of different factors, and standardizes the natural and artificial interference intensity for direct comparison, thereby accurately reflecting the abnormal degree of the current interference and further improving the timeliness and accuracy of the evaluation.

[0023] Another aspect of the application also provides a dynamic reference index calculation system for grassland ecological quality evaluation, comprising a processor, an input device, an output device and a memory, which are connected to each other, wherein the memory is used to store a computer program, the computer program comprises program instructions, and the processor is configured to invoke the program instructions to execute the dynamic reference index calculation method for grassland ecological quality evaluation according to any one of the previous aspect.

[0024] The dynamic reference index calculation system for grassland ecological quality evaluation has compact structure, stable performance, high integration and simple composition, can stably execute the dynamic reference index calculation method for grassland ecological quality evaluation according to the previous aspect, and further improves the overall applicability and practical application ability of the application. Attached Figure Description

[0025] Figure 1 This is a flowchart illustrating a method for calculating dynamic reference indicators for grassland ecological quality assessment according to an embodiment of the present invention.

[0026] Figure 2 This is a schematic diagram of a dynamic reference index calculation system for grassland ecological quality assessment according to an embodiment of the present invention. Detailed Implementation

[0027] Specific embodiments of the present invention will now be described in detail. It should be noted that the embodiments described herein are for illustrative purposes only and are not intended to limit the invention. In the following description, numerous specific details are set forth in order to provide a thorough understanding of the invention. However, it will be apparent to those skilled in the art that these specific details are not necessary to practice the invention. In other instances, well-known circuits, software, or methods have not been specifically described to avoid obscuring the invention.

[0028] Throughout this specification, references to "an embodiment," "an embodiment," "an example," or "an example" mean that a particular feature, structure, or characteristic described in connection with that embodiment or example is included in at least one embodiment of the invention. Therefore, the phrases "in an embodiment," "in an embodiment," "an example," or "an example" appearing in various places throughout the specification do not necessarily refer to the same embodiment or example. Furthermore, specific features, structures, or characteristics can be combined in one or more embodiments or examples in any suitable combination and / or sub-combination. Moreover, those skilled in the art will understand that the illustrations provided herein are for illustrative purposes and are not necessarily drawn to scale.

[0029] Please see Figure 1 To address the shortcomings of the prior art, in one alternative embodiment, such as Figure 1 The method for calculating dynamic reference indicators for grassland ecological quality assessment, as shown, includes the following steps:

[0030] Step S1: Obtain historical benchmark data of the target area covering sample plots with different levels of disturbance and no degradation, and construct a reference system time series prediction model based on the historical benchmark data.

[0031] The process of obtaining historical baseline data for non-degraded sample plots covering different levels of disturbance in the target area includes the following sub-steps:

[0032] Step S101: Obtain the original historical data of the target area covering sample plots with different levels of disturbance and without degradation.

[0033] In this embodiment, obtaining the original historical data of sample plots in the target area that cover different levels of disturbance and are not degraded specifically means selecting sample plots in the target area that cover different levels of disturbance (including different intensities of natural and human disturbance) and are determined to be non-degraded, and collecting original information including grassland subtype, historical biological indicators (original functional group species cover, original functional group species aboveground biomass, degradation indicator species cover, degradation indicator species aboveground biomass), historical environmental data, and historical disturbance data.

[0034] Step S102: Impute missing values ​​and remove outliers from the original historical data to obtain optimized historical data.

[0035] In this embodiment, for continuous data in the original historical data (such as native functional group species cover, aboveground biomass, precipitation and temperature in historical environmental data, grazing intensity in historical disturbance data, etc.), time series interpolation or similarity based on sample plots of the same grassland subtype and disturbance level can be used to fill in the gaps. If it is categorical data (such as grassland subtype), the mainstream types of neighboring sample plots can be referenced for supplementation. For outliers, they need to be identified and removed by combining statistical methods (such as extreme values ​​that deviate too much from the mean) and ecological rationality (such as cover values ​​exceeding 100%, negative biomass values, abnormally high importance values ​​of degradation indicator species in non-degraded sample plots that violate the characteristics of non-degraded sample plots, etc.).

[0036] Step S103: Normalize the optimized historical data to obtain historical baseline data.

[0037] In this embodiment, when normalizing the optimized historical data to obtain historical baseline data, it is necessary to first group the data according to grassland subtypes (such as typical grassland, desert steppe, etc.) to ensure that the data in the same group have ecological type consistency. Then, the min-max normalization method is used for different data types in each group. For historical biological indicators, the minimum and maximum values ​​of the corresponding indicators in the optimized historical data under the grassland subtype are used as the benchmark, and they are mapped to the [0,1] interval by formula to retain the relative differences of each indicator within the same subtype (such as the ranking of the original cover). For historical environmental data (such as annual precipitation, annual average temperature) and historical disturbance data (such as grazing intensity, drought index), after grouping by grassland subtype, normalization is performed based on the extreme value range of the corresponding indicators within the subtype to eliminate the numerical incomparability of different indicators due to the difference in units. The resulting historical baseline data not only retains the ecological gradient characteristics of each indicator within the same grassland subtype (such as the response relationship between disturbance intensity and biological indicators), but also achieves the standardization of different indicators.

[0038] The historical baseline data includes grassland subtypes, historical biological indicators, historical environmental data, and historical disturbance data. The historical biological indicators include native functional group species cover, native functional group aboveground biomass, degradation indicator species cover, and degradation indicator species aboveground biomass.

[0039] The construction of the reference system time series prediction model based on the historical benchmark data specifically includes the following sub-steps:

[0040] Step S111: Calculate the biomass percentage of the degraded indicator species in the sample plot using the aboveground biomass of the degraded indicator species.

[0041] In this embodiment, when calculating the biomass percentage of the degraded indicator species in the sample plot using the aboveground biomass of the degraded indicator species, it is necessary to first count the aboveground parts of all plants in the target sample plot, measure and record the aboveground biomass of the degraded indicator species, and simultaneously measure and summarize the aboveground biomass of all plants in the sample plot (including native functional group species, degraded indicator species and other associated species) to obtain the total aboveground biomass of the sample plot. Finally, the biomass percentage of the degraded indicator species in the sample plot is obtained by calculating the ratio of the aboveground biomass of the degraded indicator species to the total aboveground biomass of the sample plot.

[0042] Step S112: Calculate the coverage percentage of the degraded indicator species in the sample plot using the coverage of the degraded indicator species.

[0043] In this embodiment, the projection coverage method (i.e., the ratio of the vertical projection area of ​​the aboveground parts of the plants to the area of ​​the quadrat) is used to record the coverage of each degraded indicator species. The total coverage of degraded indicator species in the quadrat is accumulated. The total coverage of all plants in the quadrat is measured simultaneously as the total coverage of the quadrat. Finally, the coverage ratio of degraded indicator species in the quadrat is obtained by calculating the ratio of the total coverage of degraded indicator species to the total coverage of the quadrat.

[0044] Step S113: Calculate the importance value of the degradation indicator species in the sample plot based on the biomass ratio and the coverage ratio.

[0045] In this embodiment, the arithmetic mean method is used to calculate the importance value of degraded indicator species. The importance value of degraded indicator species = (biomass percentage + cover percentage) / 2. It should be noted that before the calculation, it is necessary to ensure that the two percentage data are the results of the same sampling period (such as the end of the growing season) and the same standard quadrat, and have passed the species classification verification (refer to the degraded indicator species list of the target area to avoid including non-indicator species in the calculation). If the plot contains multiple degraded indicator species, the biomass percentage and cover percentage of each species should be calculated separately first, and then the total biomass percentage and total cover percentage of the group should be accumulated before being substituted into the formula.

[0046] Step S114: Construct training samples based on the grassland subtype, the native functional group species cover, the aboveground biomass of the native functional group species, the importance value of the degradation indicator species, the historical environmental data, and the historical disturbance data.

[0047] In this embodiment, the annual data for each consecutive 5 years is used as the input window. The window includes the native functional group species cover (measured at the end of each growing season), native functional group aboveground biomass, importance value of degradation indicator species, historical environmental data (annual total precipitation, annual average temperature, sourced from the meteorological station associated with the sample plot), and historical interference factors (natural interference: annual drought index (annual precipitation / average of the past 10 years), fire frequency; human interference: annual grazing intensity (carrying capacity / theoretical carrying capacity), fence damage rate, all calibrated by field surveys and remote sensing data). At the same time, grassland subtype is included as a category feature (converted into numerical features through unique thermal encoding) in the input. The output corresponding to the window is the native functional group species cover, native functional group aboveground biomass, and importance value of degradation indicator species for the 6th year (as the prediction target of the model). The sample construction needs to cover a continuous monitoring period of at least 10 years, forming continuous sliding samples such as "t=1-5 years data → t=6 years target" and "t=2-6 years data → t=7 years target". The samples under each grassland subtype need to include gradients of different levels of interference, such as mild, moderate and severe. All indicators are normalized in the same way as historical benchmark data. The final training samples need to be associated with sample ID, grassland subtype label and timestamp to ensure that the sample size of each subtype meets the training requirements of LSTM network.

[0048] Step S115: Construct a reference system time series prediction model, and train and evaluate the reference system time series prediction model using the training samples.

[0049] The construction of the reference-based time series prediction model specifically includes the following sub-steps:

[0050] Step S11501: Statistical analysis is performed on the historical baseline data to obtain the variation patterns of native functional group species cover and aboveground biomass.

[0051] In this embodiment, the data is grouped by grassland subtype (ensuring consistency in ecological type within the same group), and the native functional group species cover is calculated for at least 10 consecutive years within each group. and aboveground biomass Time-series data are used to calculate the changes between adjacent years, and then multi-dimensional analysis is carried out: the consistency ratio of the change direction of the two under different disturbance gradients (mild / moderate / severe natural and anthropogenic disturbances) is statistically analyzed (the proportion of samples with the same increase or decrease should be ≥80% to reflect the coordinated trend), and samples with contradictory change directions are screened out. The distribution characteristics of the sum of the two absolute values ​​of change in these contradictory samples (absolute value is due to the change amount being positive or negative) are calculated (such as the maximum value or 95th percentile, etc.). Finally, the core law that the species cover of the original functional group and the aboveground biomass increase or decrease together in most cases, and only show small-scale contradictions in a few natural fluctuations are determined. At the same time, the historical resilience threshold is determined (taking the maximum value of the sum of absolute values ​​in the historical contradictory samples).

[0052] Step S11502: Set collaborative constraint terms based on the change pattern.

[0053] The collaborative constraint terms satisfy the following formula:

[0054]

[0055] in, For collaborative constraint terms, For the sample size, For indicator functions, For the predicted first Changes in native functional group species coverage of individual samples The direction changes, increasing to positive and decreasing to negative. For the predicted first Changes in biomass of primary functional groups in each sample The direction changes, increasing to positive and decreasing to negative. The historical resilience threshold is obtained through statistical analysis of historical benchmark data.

[0056] The above formula uses an indicator function. Identify the changes in species cover of native functional groups in the model prediction and aboveground biomass change The changing trends are contradictory (e.g., predicted increase in cover but predicted decrease in biomass), and then through... Screening and penalizing the magnitude of contradictions, historical resilience threshold. It is statistically derived from historical baseline data and represents the maximum permissible magnitude of contradictions under natural fluctuations, such as small-scale trend deviations caused by short-term droughts. Penalties are only triggered when the magnitude of the contradictions (the sum of the absolute values ​​of the changes in both) exceeds this threshold, and the intensity of the penalty increases linearly with the magnitude of the excess. This design respects the resilience and tolerance of grassland ecosystems to short-term disturbances (no penalty for small contradictions) while strictly constraining drastic contradictions that violate the core synergistic laws (such as a sudden increase in cover but a sudden decrease in biomass), ensuring that the dynamics of native functional groups predicted by the model conform to the ecological evolution logic of long-term co-increase and co-decrease and short-term limited fluctuations.

[0057] It should be noted that, with For example, it means the first The difference between the predicted value and the actual value of a sample.

[0058] Step S11503: Add the collaborative constraint term to the loss function constructed with mean square error to obtain the optimized loss function.

[0059]

[0060] in, To optimize the loss function, Let the mean squared error loss function be . Penalty coefficient These are collaborative constraint terms.

[0061] The penalty coefficient in the above formula It can be optimized through test set. If the model's prediction accuracy is sufficient but the ecological logic deviates, the accuracy can be increased. Strengthening collaborative constraints and prioritizing fitting when data noise is high can reduce [the impact of noise]. .

[0062] Step S11504: Construct a reference time series prediction model based on the optimized loss function.

[0063] In this embodiment, a Long Short-Term Memory (LSTM) network is selected to adapt to the temporal evolution dependence of grassland ecological indicators. The loss function is optimized during model training, and the model parameters are shaped by a dual-track regulation of accuracy fitting and ecological constraints. On the one hand, the mean squared error term drives the model to learn the numerical patterns of historical data, minimizes the deviation between the predicted values ​​of native functional group cover and biomass and the actual values, and ensures the quantitative accuracy of temporal prediction. On the other hand, the ecological coordination constraint term constructs a penalty logic of trend contradiction and amplitude threshold. When the predicted cover and biomass change direction conflict (such as cover increasing while biomass decreases), and the sum of the absolute values ​​of the two changes exceeds the ecological resilience threshold of historical statistics (representing the maximum contradiction amplitude allowed by natural fluctuations), a penalty is imposed on the sample, forcing the model to avoid violating the prediction pattern of long-term coordination between cover and biomass.

[0064] Finally, the dataset was stratified into training and test sets according to grassland subtypes. The training set was used to train the time-series prediction model for the reference system. The parameters were updated by backpropagation through optimization of the loss function until the model converged. In the evaluation phase, the model was validated on the test set using dual indicators: quantitatively, the mean squared error of the prediction of native functional group cover, aboveground biomass, and importance of degradation indicator species was calculated (it must be less than 15% of the standard deviation of historical baseline data); qualitatively, the proportion of statistically contradictory samples (≤5%) and the magnitude of the contradiction (both ≤ the historical resilience threshold ε) were measured. At the same time, it was ensured that the test results of each grassland subtype had no significant bias (the error between subtypes was <20%). Finally, it was determined whether the model possessed the accuracy of time-series prediction and ecological logic consistency.

[0065] Step S2: Based on the historical monitoring data of the grassland to be evaluated obtained in advance, the potential value of the dynamic reference index is obtained using the reference system time series prediction model.

[0066] In this embodiment, historical monitoring data of the grassland to be evaluated for the previous 5 consecutive years are extracted, including biological indicators, annual environmental data, grassland subtype, and disturbance data. These data are then normalized to ensure consistency with the training data processing method and the input label dimension of the training data. Subsequently, the data is input into a pre-trained reference system time series prediction model. By learning the evolution pattern of the grassland to be evaluated data over the previous 5 years, the model outputs the potential values ​​of the original functional group species indicators (coverage, aboveground biomass) and the potential values ​​of the degradation indicator species importance in the target year of the grassland to be evaluated.

[0067] Step S3: Extract interference factors from the historical benchmark data and set interference weights for the interference factors.

[0068] The process of extracting interference factors from the historical benchmark data and setting interference weights for these interference factors specifically includes the following sub-steps:

[0069] Step S301: Extract historical reference interference data from the historical reference data, and determine the interference factor based on the historical reference interference data.

[0070] In this embodiment, historical baseline disturbance data is first selected from historical baseline data covering sample plots with different levels of disturbance and in a non-degraded or stable state. The data belonging to the disturbance category are then selected as historical baseline disturbance data. Natural disturbance-related data include the annual drought index (calculated as "annual precipitation / average of the past 10 years") and fire frequency (the number of fires that occur within the year). Human disturbance-related data include the annual grazing intensity (calculated as "livestock carrying capacity / theoretical carrying capacity") and fence damage rate (the proportion of fence damage length to total length within the year). Finally, the natural disturbance indicators (drought index, fire frequency) and human disturbance indicators (grazing intensity, fence damage rate) that have been screened and verified are determined as disturbance factors to ensure that they can comprehensively reflect the intensity of natural and human disturbances that have historically affected grassland ecology.

[0071] Step S302: Calculate the annual comprehensive change rate of the original functional group species using the historical benchmark data.

[0072] The calculation of the annual comprehensive change rate of the original functional groups using the historical benchmark data specifically includes the following sub-steps:

[0073] Step S30201: Extract the annual change rate set of native functional group species cover and the annual change rate set of native functional group species aboveground biomass from the historical baseline disturbance data.

[0074] In this embodiment, the historical baseline disturbance data is grouped by grassland subtype, and the annual cover data and annual aboveground biomass data of the native functional groups continuously monitored in each group are extracted. The change rate of adjacent years is calculated (the difference between the value of the following year and the value of the previous year divided by the value of the previous year), thereby forming the set of annual change rate of native functional group cover and the set of annual change rate of aboveground biomass corresponding to each grassland subtype.

[0075] Step S30202: Calculate the mean annual change rate of cover, the standard deviation of annual change rate of cover, the mean annual change rate of biomass, and the standard deviation of annual change rate of biomass based on the set of annual change rates of cover of the original functional group species and the set of annual change rates of aboveground biomass of the original functional group species.

[0076] In this embodiment, after grouping by grassland subtype, for the set of annual change rates of native functional group species cover in each group, the arithmetic mean of all annual change rates (i.e., the mean of annual change rate of cover) and the standard deviation reflecting the dispersion of the data (i.e., the standard deviation of annual change rate of cover) are calculated. Similarly, for the set of annual change rates of aboveground biomass of native functional group species, the arithmetic mean of all annual change rates (i.e., the mean of annual change rate of biomass) and the standard deviation (i.e., the standard deviation of annual change rate of biomass) are calculated respectively.

[0077] Step S30203: Calculate cover stability using the mean annual cover change rate and the standard deviation of the annual cover change rate, and calculate biomass stability using the mean annual biomass change rate and the standard deviation of the annual biomass change rate.

[0078] Coverage stability satisfies the following formula:

[0079]

[0080] Biomass stability satisfies the following formula:

[0081]

[0082] in, For coverage stability, The standard deviation of the annual rate of change of coverage. The average annual rate of change of coverage. For biomass stability, The standard deviation of the annual rate of change of biomass. This represents the average annual rate of change in biomass.

[0083] In the above formula, the standard deviation of the annual change rate of coverage represents the fluctuation range of interannual coverage, the absolute value of the mean of the annual change rate of coverage represents the average trend strength of interannual coverage, and the ratio of the two, coverage stability, represents the proportion of coverage fluctuation range to its average trend. The larger the ratio, the more the interannual coverage fluctuation deviates from the average trend, and the worse the dynamic stability of functional group coverage (such as a sudden drop in coverage in drought years, exceeding the normal interannual fluctuation range). Similarly, biomass stability measures the proportion of interannual fluctuation of aboveground biomass of primary functional groups to its average trend.

[0084] Step S30204: Calculate the cover weight of native functional group species and the aboveground biomass weight of native functional group species using the cover stability and biomass stability, respectively.

[0085] In this embodiment, the lower the stability, the smoother the interannual dynamics of the indicator; conversely, the more drastic the indicator's response to the environment / disturbance, i.e., the higher its sensitivity. Since changes in sensitive indicators can better reflect the degradation trend of grassland functional groups, weights are constructed by normalizing the inverse of stability.

[0086] The native functional group coverage weights satisfy the following formula:

[0087]

[0088] The aboveground biomass weights of native functional groups satisfy the following formula:

[0089]

[0090] Substituting the cover stability formula and the biomass stability formula into the above weighting formula, we get:

[0091]

[0092]

[0093] in, For the native functional group, the coverage weight, The aboveground biomass weight of the original functional group species. For coverage stability, For biomass stability, The standard deviation of the annual rate of change of coverage. The average annual rate of change of coverage. The standard deviation of the annual rate of change of biomass. This represents the average annual rate of change in biomass.

[0094] Step S30205: Calculate the annual comprehensive change rate of the original functional group species using the historical benchmark data based on the original functional group species coverage weight and the original functional group species aboveground biomass weight.

[0095] The annual comprehensive change rate of the original functional group of species satisfies the following formula:

[0096]

[0097] in, for Annual comprehensive change rate of the original functional group species For the native functional group, the coverage weight, The aboveground biomass weight of the original functional group species. for Annual native functional group species coverage for Annual aboveground biomass of native functional groups for Annual native functional group species coverage for Annual aboveground biomass of native functional groups.

[0098] Step S303: Train a random forest change rate prediction model using the historical baseline disturbance data and the annual comprehensive change rate of the original functional groups.

[0099] In this embodiment, after normalizing the historical baseline disturbance data and the annual comprehensive change rate of the original functional groups, a sample set is formed. The random forest change rate prediction model is trained using the historical baseline disturbance data as the input feature and the annual comprehensive change rate of the original functional groups as the output feature.

[0100] Step S304: Calculate the interference weight of the interference factor based on the random forest rate of change prediction model.

[0101] The calculation of the interference weights of the interference factors based on the random forest rate of change prediction model specifically includes the following sub-steps:

[0102] Step S30401: Extract the node information of the interference factor that participates in node splitting in the random forest change rate prediction model.

[0103] In this embodiment, it is necessary to traverse all nodes of each decision tree in the model, filter out nodes whose splits are dominated by interference factors (annual drought index and fire frequency in natural interferences, and annual grazing intensity and fence damage rate in human interferences), and record their detailed information one by one: including node attributes (the node's number in its decision tree and the corresponding decision tree number), splitting factor (i.e., the specific interference factor on which the node splits samples), parent node characteristics (the total number of samples contained in the parent node, and the comprehensive change rate of the original functional groups of all samples in the parent node), left child node characteristics (the number of samples in the left child node after splitting by the interference factor, and the comprehensive change rate of the original functional groups of all samples in the left child node), and right child node characteristics (the number of samples in the right child node, and the comprehensive change rate of the original functional groups of all samples in the right child node), so as to fully capture the specific details of the interference factors participating in node splits during the model decision-making process.

[0104] Step S30402: Calculate the mean square error reduction value of the split node dominated by the interference factor based on the node information.

[0105] The mean squared error decrease value satisfies:

[0106]

[0107] in, For the first The first decision tree Each node is composed of interference factors. The decrease in the mean square error of the annual comprehensive change rate of the original functional group after the split. For the first The first decision tree The mean square error of the annual comprehensive change rate of the original functional group of each node. For the first The first decision tree Each node is composed of interference factors. The mean square error of the annual composite change rate of the original functional group of the left child node after splitting. For the first The first decision tree Each node is composed of interference factors. The mean square error of the annual composite change rate of the original functional group of the right child node after splitting. For the first The first decision tree The number of samples per node. For the first The first decision tree Each node is composed of interference factors. The number of samples in the left child node after the split. For the first The first decision tree Each node is composed of interference factors. The number of samples in the right child node after splitting.

[0108]

[0109] in, For the first The annual composite change rate of the original functional groups of each sample For the first The first decision tree The average annual comprehensive change rate of the original functional groups of each node.

[0110] The core of decision tree node splitting is to divide samples by features (interference factors) to make the differences in the annual comprehensive change rate of the original functional groups within the child nodes smaller (mean squared error reduced). The calculation of the decrease in mean squared error is essentially to quantify the difference between the error of the parent node and the weighted error of the child node when a certain interference factor splits the node. The larger the difference, the stronger the contribution of the interference factor to distinguishing the heterogeneity of the comprehensive change rate (the change rate within the child nodes after splitting is more similar, and the prediction is more accurate).

[0111] From an ecological perspective, if a disturbance factor (such as grazing intensity) causes a large decrease in the mean squared error when a node splits, it means that this factor can significantly distinguish between sample groups with high and low overall change rates (e.g., after a node with high grazing intensity splits, the change rate of the left child node drops sharply, while the right child node remains relatively stable). This indicates that grazing intensity is a key disturbance driving changes in functional groups. Conversely, a small decrease in the mean squared error indicates that the factor has a weak ability to distinguish change rates. Therefore, by analyzing the decrease in the mean squared error, we can inversely identify which disturbance factors, at which nodes, have a stronger explanatory power for the dynamics of functional groups during the model's decision-making process.

[0112] Step S30403: Calculate the total importance of the interference factors in the random forest model using the mean squared error decrease value.

[0113] In this embodiment, all decision trees in the model are traversed, and all nodes that dominate the splitting of each interference factor are selected. The mean squared error reduction values ​​of these nodes are enumerated one by one and accumulated to obtain the original cumulative importance value of the interference factor, which is the total importance of the interference factor.

[0114] Total importance satisfies the following formula:

[0115]

[0116] in, Interference factor The overall importance, The number of decision trees in the random forest model. For the first The decision tree contains interference factors The number of nodes that dominate the split. For the first The first decision tree Each node is composed of interference factors. The decrease in the mean square error of the annual comprehensive change rate of the original functional group after splitting.

[0117] Step S30404: Calculate the interference weight of the interference factor based on the total importance.

[0118] In this embodiment, the original cumulative values ​​of all interference factors are summed, and the sum is normalized with the cumulative value of each individual factor as the numerator and the sum as the denominator, so as to obtain the total importance ratio of each interference factor, i.e., the interference weight.

[0119] The interference weights satisfy the following formula:

[0120]

[0121] in, Interference factor The weight, This represents the number of different types of interference factors. For the first The total importance of each interfering factor.

[0122] Step S4: Based on the historical monitoring data and the historical benchmark data, the potential value of the dynamic reference index is corrected using the interference weight to obtain the dynamic reference index.

[0123] The process of correcting the potential value of the dynamic reference indicator based on the historical monitoring data and the historical benchmark data using the interference weight to obtain the dynamic reference indicator specifically includes the following sub-steps:

[0124] Step S401: Calculate the current normalized strength of the interference factor using the historical monitoring data and the historical benchmark data.

[0125] The calculation of the current standardized strength of the interference factor using the historical monitoring data and the historical benchmark data specifically includes the following steps:

[0126] Step S40101: Extract the monitoring interference data for the most recent year from the historical monitoring data.

[0127] In this embodiment, the historical monitoring data comes from the data of the most recent 5 years, and the interference data of the most recent year is extracted from the historical monitoring data to obtain the monitoring interference data.

[0128] Step S40102: Calculate the historical interference mean and historical interference standard deviation based on the historical benchmark data.

[0129] In this embodiment, the corresponding data for all monitoring years are extracted from the historical baseline data according to the factors of natural disturbance (drought index, fire frequency) and human disturbance (grazing intensity, fence damage rate). For each disturbance factor, the arithmetic mean of all annual data (i.e. the historical disturbance mean of the factor, reflecting the average intensity) and standard deviation (i.e. the historical disturbance standard deviation of the factor, reflecting the degree of interannual fluctuation) are calculated.

[0130] Step S40103: Calculate the current standardized strength of the interference factor based on the historical interference mean, the historical interference standard deviation, and the monitored interference data.

[0131] The current standardization strength satisfies the following formula:

[0132]

[0133] in, For the first The current normalized strength of the interference factors, No. Measured values ​​of various interference factors For the first Historical interference mean of various interference factors For the first The historical standard deviation of interference factors.

[0134] The current standardized intensity converts the measured intensity of the interference factor into a standard deviation multiple relative to the historical average. The positive or negative value reflects that the interference is stronger or weaker than the historical average, and the absolute value reflects the degree of deviation from historical fluctuations. Its core significance is to eliminate the dimensional differences between different interference factors, allowing the intensity of natural and man-made interferences to be compared horizontally. At the same time, it quantifies the significance of the current interference deviating from the normal state through the historical standard deviation. This not only solves the comparability problem of multiple interference factors, but also allows us to judge the degree of abnormality of the current interference from the perspective of historical fluctuations.

[0135] Step S402: Calculate the total impact magnitude of the current interference using the current normalization intensity and the interference weight.

[0136] The total impact of the current interference satisfies the following formula:

[0137]

[0138] in, This represents the total impact of the current interference. The number of interference factor types. For the first The weights of various interference factors No. Measured values ​​of various interference factors For the first Historical interference mean of various interference factors For the first The historical standard deviation of interference factors.

[0139] The above formula uses the ecological impact of disturbance (weight) Using as the scale, the standardized intensity of each interference factor deviating from the historical normal is aggregated. ), quantifying the total driving effect of the current disturbance combination on the annual comprehensive change rate of the original functional group.

[0140] Step S403: Based on the current total impact of interference, the potential value of the dynamic reference index is corrected to obtain the dynamic reference index.

[0141] In this embodiment, the dynamic reference indicators include dynamic reference values ​​for native functional group species indicators and dynamic reference values ​​for the importance of degraded indicator species. A two-way correction logic is designed around the ecological antagonistic relationship between native functional groups and degraded indicator species. For native functional group indicators, the potential value of the native functional group species indicator reference system represents the ideal state under historical average disturbance. When the total impact of disturbance is positive (the current disturbance combination as a whole promotes the native functional group), it indicates that the current ecology is more conducive to the development of the native functional group. Therefore, through... Amplify the potential value so that the dynamic reference value is higher than the historical baseline. If the total impact of the disturbance is negative (the disturbance as a whole suppresses the original disturbance), then... Narrowing the potential value aligns with the pattern of interference suppression—the original state being worse.

[0142] For the importance value of degradation indicator species, the potential value of its reference system represents the degradation baseline under historical average disturbance, which is usually lower and corresponds to ecological health. Since the response of degradation indicator species is inverse to that of native functional groups (disturbance promotes native populations while inhibiting degradation, and disturbance inhibits native populations while promoting degradation), therefore, a [reference system is used]. Correction: The total impact of the current interference is greater than 0 (interference promotes the original state, i.e., inhibits degradation). It will reduce the potential value and make the degradation less severe, which is consistent with the logic that interference promotes the original state. The current total impact of interference is less than 0 (interference inhibits the original state, that is, it promotes degradation). It will amplify the potential value, making the degradation more significant, which aligns with the original pattern of interference suppression.

[0143] The revised dynamic reference index is no longer a historical static benchmark detached from the current disturbance, but a dynamic expected value that integrates the comprehensive effect of the current disturbance with the antagonistic relationship of ecological indicators. It provides a more realistic reference scale for the subsequent assessment of grassland ecological status (such as whether the original functional groups are better than expected and whether the degradation exceeds the threshold).

[0144] The dynamic reference index satisfies the following formula:

[0145]

[0146]

[0147] in This serves as a dynamic reference value for the original functional group's indicators. As a dynamic reference value for important values ​​of degenerate indicator species, As a reference system for the original functional group species index potential value, Potential values ​​serve as important references for degenerate indicator species.

[0148] After obtaining the dynamic reference indicators, a specific assessment can be carried out in conjunction with the grassland ecological quality assessment system. Using the dynamic reference indicators as a benchmark, the comprehensive change rate of the original functional groups of the grassland to be evaluated, as well as the importance value of degradation indicator species, should be calculated first.

[0149] Based on the grading standards, the comprehensive change rate of native functional groups is divided into five levels: >120, 105-120, 95-105, 80-95, and ≤80, corresponding to scores of 5, 4, 3, 2, and 1, respectively. The importance value of degraded indicator species is divided into five levels: <5, 5-15, 15-30, 30-50, and ≥50, corresponding to scores of 5, 4, 3, 2, and 1, respectively. The Ecological Quality Index (EQI) is then calculated with a weight of 0.7 for the comprehensive change rate of native functional groups and a weight of 0.3 for the importance value of degraded indicator species. Finally, based on the EQI grading: >4.5 is "Excellent," 3.5-4.5 is "Good," 2.5-3.5 is "Moderate," 1.5-2.5 is "Poor," and ≤1.5 is "Very Poor," achieving a precise evaluation of grassland ecological quality. Because this process incorporates real-time disturbances and environmental changes into the dynamic reference indicators, the evaluation results are more closely aligned with the actual ecological condition of the grassland, providing a scientific basis for protection and restoration measures.

[0150] The calculation method of the ecological quality index is as follows: the scores of the two evaluation indicators, namely the comprehensive change rate of native functional groups (weight 0.7) and the importance value of degraded indicator species (weight 0.3), are obtained by multiplying them by their respective weights and summing them.

[0151] like Figure 2As shown, in another aspect, the present invention also provides a dynamic reference index calculation system for grassland ecological quality assessment, comprising: a processor, an input device, an output device, and a memory, wherein the processor, the input device, the output device, and the memory are interconnected, wherein the memory is used to store a computer program, the computer program includes program instructions, and the processor is configured to call the program instructions to execute the relevant steps of the relevant embodiments of the dynamic reference index calculation method for grassland ecological quality assessment of the present invention.

[0152] This invention provides a dynamic reference index calculation system for grassland ecological quality assessment. The functional components can be integrated into a single processing unit, exist as individual physical entities, or be integrated into a single unit. These integrated components can be implemented in hardware or software.

[0153] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention, and they should all be covered within the scope of the claims and specification of the present invention.

Claims

1. A method for calculating dynamic reference indicators for grassland ecological quality assessment, characterized in that, The method includes: Historical baseline data of non-degraded sample plots covering different levels of disturbance in the target area are obtained, and a reference system time-series prediction model is constructed based on the historical baseline data. The construction of the reference system time-series prediction model includes: performing statistical analysis on the historical baseline data to obtain the variation patterns of native functional group species cover and aboveground biomass; setting collaborative constraint terms based on the variation patterns; adding the collaborative constraint terms to a loss function constructed using mean squared error to obtain an optimized loss function; and constructing the reference system time-series prediction model based on the optimized loss function. Based on the historical monitoring data of the grassland to be evaluated obtained in advance, the potential value of the dynamic reference index is obtained using the reference system time series prediction model. Extracting interference factors from the historical baseline data and assigning interference weights to the interference factors includes: extracting historical baseline interference data from the historical baseline data and determining interference factors based on the historical baseline interference data; calculating the annual comprehensive change rate of native functional groups using the historical baseline data; training a random forest change rate prediction model using the historical baseline interference data and the annual comprehensive change rate of native functional groups; and calculating the interference weights of the interference factors based on the random forest change rate prediction model. The step of calculating the annual comprehensive change rate of native functional group species using the historical benchmark data includes: extracting the annual change rate set of native functional group species cover and the annual change rate set of native functional group species aboveground biomass from the historical benchmark interference data; calculating the mean, standard deviation, mean, and standard deviation of the annual change rate of cover, biomass, and annual change rate of biomass, respectively, based on the annual change rate set of native functional group species cover and the annual change rate set of native functional group species aboveground biomass; calculating cover stability using the mean and standard deviation of the annual change rate of cover, and calculating biomass stability using the mean and standard deviation of the annual change rate of biomass; calculating the native functional group species cover weight and native functional group species aboveground biomass weight, respectively, using the cover stability and biomass stability; and calculating the annual comprehensive change rate of native functional group species using the historical benchmark data based on the native functional group species cover weight and native functional group species aboveground biomass weight. Based on the historical monitoring data and the historical benchmark data, the potential value of the dynamic reference indicator is corrected using the interference weight to obtain the dynamic reference indicator.

2. The method for calculating dynamic reference indicators for grassland ecological quality assessment according to claim 1, characterized in that, The acquisition of historical baseline data for the target area, covering sample plots with different levels of disturbance and no degradation, includes: Obtain raw historical data of the target area covering sample plots with different levels of disturbance and no degradation; The original historical data is optimized by imputing missing values ​​and removing outliers. The optimized historical data is normalized to obtain historical baseline data.

3. The method for calculating dynamic reference indicators for grassland ecological quality assessment according to claim 1, characterized in that, The historical baseline data includes grassland subtypes, historical biological indicators, historical environmental data, and historical disturbance data. The historical biological indicators include native functional group species cover, native functional group aboveground biomass, degradation indicator species cover, and degradation indicator species aboveground biomass. The construction of a reference system time-series prediction model based on the historical baseline data includes: The aboveground biomass of the degraded indicator species was used to calculate the proportion of biomass of the degraded indicator species in the sample plot; The coverage percentage of the degraded indicator species in the sample plot is calculated using the coverage of the degraded indicator species. The importance value of the degraded indicator species in the sample plot is calculated based on the biomass ratio and the cover ratio. Training samples are constructed based on the grassland subtype, the native functional group species cover, the native functional group species aboveground biomass, the degradation indicator species importance value, the historical environmental data, and the historical disturbance data. A reference system time series prediction model is constructed, and the reference system time series prediction model is trained and evaluated using the training samples.

4. The method for calculating dynamic reference indicators for grassland ecological quality assessment according to claim 1, characterized in that, The calculation of the interference weights of the interference factors based on the random forest rate of change prediction model includes: Extract the node information of the interference factor that participates in node splitting in the random forest rate of change prediction model; The mean square error reduction value of the split node dominated by the interference factor is calculated based on the node information. The total importance of the interference factors in the random forest model is calculated using the mean squared error decrease value. The interference weight of the interference factor is calculated based on the total importance.

5. The method for calculating dynamic reference indicators for grassland ecological quality assessment according to claim 1, characterized in that, The process of correcting the potential value of the dynamic reference indicator based on the historical monitoring data and the historical benchmark data using the interference weight to obtain the dynamic reference indicator includes: The current normalized strength of the interference factor is calculated using the historical monitoring data and the historical benchmark data; The total impact magnitude of the current interference is calculated using the current standardization strength and the interference weight; The dynamic reference index is obtained by correcting the potential value of the dynamic reference index based on the current total impact of the interference.

6. The method for calculating dynamic reference indicators for grassland ecological quality assessment according to claim 5, characterized in that, The calculation of the current standardized strength of the interference factor using the historical monitoring data and the historical benchmark data includes: Extract the monitoring interference data for the most recent year from the historical monitoring data; Calculate the historical interference mean and historical interference standard deviation based on the historical benchmark data; The current standardized strength of the interference factor is calculated based on the historical interference mean, the historical interference standard deviation, and the monitored interference data.

7. A dynamic reference index calculation system for grassland ecological quality assessment, characterized in that, include: The system includes a processor, an input device, an output device, and a memory, all interconnected. The memory stores a computer program, which includes program instructions. The processor is configured to invoke the program instructions to execute a dynamic reference index calculation method for grassland ecological quality assessment as described in any one of claims 1 to 6.

Images